Single-particle reconstruction (SPR) electron microscopy is a powerful technique for obtaining the 3D structure of macromolecular complexes in near-native states. A major problem in SPR is that flexibility gives rise to multiple conformations when the macromolecular particles are frozen in vitreous ice. Traditional SPR assumes that macromolecular particles are rigid, and gives good reconstructions only when all particles have the identical structure. We propose to develop new reconstruction methods which eliminate the rigidity assumption altogether. These methods can reconstruct a continuously flexible structure. Two classes of methods are proposed: (a) methods for exploratory analysis which detect the presence of structural conformational change and objectively determine the most dominant modes of flexibility, and (b) post- exploratory flexible particle reconstruction methods which reconstruct in detail the identified modes of flexibility. These methods are based on probability theory and preliminary results with actual data show that they offer significantly higher signal-to-noise ratio (SNR) reconstructions even when the data SNR is as low as -25db. Further, our methods provide a simple interpretation of particle flexibility. We will develop these algorithms and write software modules for incorporation in the SPARX cryo-EM processing environment. As a first application, we will use the software to process negative-staining and cryo-EM images of the human Dicer RNA-processing enzyme. As positive and negative controls, we will also reconstruct the structure of RNA Polymerase II and GroEL.